Local and global topological complexity measures OF ReLU neural network functions

04/12/2022

∙

by   J. Elisenda Grigsby, et al.

∙

0

∙

share

We apply a generalized piecewise-linear (PL) version of Morse theory due to Grunert-Kuhnel-Rote to define and study new local and global notions of topological complexity for fully-connected feedforward ReLU neural network functions, F: R^n -> R. Along the way, we show how to construct, for each such F, a canonical polytopal complex K(F) and a deformation retract of the domain onto K(F), yielding a convenient compact model for performing calculations. We also give a combinatorial description of local complexity for depth 2 networks, and a construction showing that local complexity can be arbitrarily high.